Best to stick with liquid securities. If ADV is under 200k shares, that's a big danger sign. Real-world performance on illiquid securities is likely even worse than modeled by Quantopian. Also, Quantopian really isn't designed to model small amounts of trading capital (e.g. $500).

Try NUGT if you want leveraged gold exposure (though it's miners instead of the metal itself) or else just increase your exposure to GLD and IAU (e.g. instead of 0.25, order 0.5) if you simply want more exposure.

I think most of the gold ETFs should track decently enough -- I can't imagine there'll be much advantage to diversifying across them. Stick to the most liquid. For shares, IAU is probably most liquid and has the lowest fee. (GLD has more liquid options though.)

This strategy creates a lot of churn, so it loses a lot of money to frictional costs (fees and spread/slippage).

One thing I'll sometimes do is test a strategy with commissions and slippage set to 0 just to see if there is any alpha at all. If there is and it's strong enough then I can focus on how to execute it in such a way that is survives slippage. Do you have any evidence that holding gold only overnight offers any advantage over buy-and-hold?

Thanks to you Viridian Hawk! I actually combined your suggestions and now it works fine. I got rid of the slippage factor and only use NUGT and IAU. Returns are nearly 200%. I know it is obviously estimated under unrealistic conditions but that is ok for my purpose.

Note: there is indeed evidence for significantly asymmetric pricings of gold related assets.